How Candid Home Inspections became a testing ground for AI-enabled operations.
Candid Home Inspections is the firm Grant Blackwell ran when he was drowning. Phones rang during crawlspaces. Reports waited at nine at night. Office help and a virtual assistant both reduced the load, but the handoffs still broke. Candid became the place to build and test better workflows. Some are in use, others are still being refined, and results are validated one workflow at a time.
Three inspectors. One owner. Too many tabs.
Candid is a small Upstate SC home inspection firm: a steady book of business, multiple licensed inspectors, and Grant as the operator. The work itself was fine. The work around the work was the problem.
- Phones rang during crawlspaces.The inspectors couldn't answer. The owner couldn't always answer. Each missed call was a potential job lost or a closing attorney unhappy.
- Emails piled up between jobs.By the time an inspector got back to the truck, three buyers and two agents had emailed. CL-100 termite reports arrived from outside labs at nine p.m.
- Reports moved by hand.Every report had to be uploaded to the right inspection record in Spectora before it could be sent to clients. Easy to miss. Easy to send late.
- Two hires helped, neither solved it.An office manager and a virtual assistant. Both took real load off. Neither was around at nine p.m. when the reports arrived. Both cost real money.
Six workflows. Built and tested in stages.
Built over several months, each workflow targets a specific operational leak. They are not all at the same maturity: some support day-to-day work, some remain controlled tests, and none is presented here as a guaranteed client result.
AI receptionist on the main line.
A voice-agent workflow designed to answer after-hours calls, use approved service information, capture the reason for the call, and hand off to a human on sensitive terms. Booking and production routing are enabled only after testing and approval.
A bridge between Sarah and Spectora.
A separate integration layer is being tested between the voice workflow and Spectora. It validates data and limits what can be written before any booking action reaches production.
Four inbox monitors.
Inbox-monitor workflows classify CL-100 letters, indoor-air-quality reports, ShowingTime confirmations, and water-quality reports, then prepare the information for the right inspection record with human review where confidence is low.
A lead-finding agent.
Researches approved public sources, scores potential opportunities against defined criteria, and prepares a review queue. A person decides what is worth pursuing.
A competitive-positioning agent.
Tracks public market signals such as service changes, review activity, and positioning, then drafts a weekly summary for the owner to review.
A five-p.m. daily digest.
Combines selected operational signals into one reviewable daily summary: bookings, workload, exceptions, and overdue items. The goal is a faster check-in, not autonomous management.
Measure first. Make the bigger claim later.
Candid's workflows are being evaluated against baseline hours, error rates, response time, and operating cost. This page documents the build and the measurement approach; it does not claim that the business is fully autonomous or that a human role has been eliminated.
This became OffLoad AI.
Candid shaped the method OffLoad AI now brings to other small businesses: map the real workflow, find the narrowest useful intervention, build with guardrails, test in stages, and measure before expanding. Different industries have different details, so every recommendation starts with the client's own process and records.
Want to map your own workflow?
Use the calculator for a rough planning estimate, then take the free Workflow Audit. If the fit is sound, the paid $500 Diagnostic validates the baseline before anything is built.